- Determine what percentage of your total traffic is eligible to be in your experiment
- Determine what percentage of your experiment traffic sees which variation
- Set a winning variation live for all of your visitors
Traffic allocation in a nutshell
Traffic allocation helps you control how much of your site traffic will be shown the experiment and each variation. For example, you might want to show the experiment to 80% of visitors to your site, and divide those visitors so that 50% get the original experience and 50% get the new variation. 20% won't be included in the experiment at all.
- Set traffic allocation to 100% to collect data as quickly as possible
- Let your experiment run for at least a week to account for fluctuations in traffic
- Use traffic allocation to quickly set a winning variation live for all visitors
What to watch out for
- Start high-risk experiments that impact revenue at a lower traffic allocation to mitigate risk
- Don't change the split between variations after you start the experiment
- Changes in traffic allocation only affect future visitors. Visitors who were already cookied into the experiment will continue to see the variation they were bucketed into.
- The actual traffic split expresses a probability and won't match the allocation on your Results page exactly. As more visitors enter the experiment, the actual split will start to match your allocation.
Traffic allocation allows you to specify the percentage of visitors who will be included in an experiment and what percentage of those visitors will be shown each variation. Want to see how to set your traffic allocation? Watch this short video.
By default, an experiment will run for 100% of your visitors and the traffic will be evenly distributed between variations. For example, an experiment with 3 variations plus the Original will be automatically set to direct 25% of the traffic to each version.
In general, we recommend that you leave traffic allocation at the default levels, because external effects on traffic or conversions will disproportionately affect the variation with less traffic. Here are a couple of reasons why you might want to change traffic allocation:
- You're running an experiment that involves a high level of risk, and you want to mitigate that risk by lowering the amount of traffic involved. We recommend that you do this by lowering total experiment traffic, not by allocating traffic unequally to variations.
- You want to send all traffic to one "winning" variation. We show you how to do this below.
Allocate traffic by clicking Options > Traffic Allocation in the Editor.
You can also allocate traffic directly from the Home page. Click on an experiment name and the details sidebar on the right will populate. Click the blue Edit text next to Traffic Allocation.
Why doesn't your experiment visitor breakdown match your traffic percentage split exactly? When you specify a visitor percentage split, you're actually specifying the probability that any particular visitor will be placed in that particular variation.
So, if you have 4 variations with traffic split equally between them, every new visitor in your experiment has a 25.00% chance of being placed in each one of these buckets. As a result the number of visitors in each bucket may not line up exactly with the percentage splits you've specified. However, over time, the number of visitors in each variation should approach 25.00% of the total.
Change the percentage of traffic included in your experiment
Moving the slider to the left will decrease the amount of traffic included in the experiment. If you change the experiment to run for only 50% of the traffic, this means that half of the visitors coming to your page will not see the experiment at all and their actions will not be included in the experiment results. The other 50% of the traffic will be included in the experiment and will be evenly distributed across the original page and the variations. You can allocate anywhere between 0.01% and 100% of traffic to your experiment.
Changes made to the overall experiment traffic allocation will only affect new users. This means existing users (whether or not they were bucketed in a variation) will keep seeing the same variation even if your change the traffic allocation settings. This also means that visitors excluded from the test will remain excluded, even if you change the overall traffic allocation to 100%.